Often when doing cross-distribution benchmarks, readers often comment on the performance of Clear Linux particularly for video encoding use-cases as surprisingly different from other distributions. Some argue that it’s just over the default CPU frequency scaling governor or compiler flag defaults, so here is a look at that with Ubuntu 19.10 daily benchmarked against Clear Linux.
On the same Core i9 9900K system I recently ran some benchmarks looking at Clear Linux vs. Ubuntu 19.10 and then Ubuntu 19.10 with various common tunables to make it more akin to Clear Linux. Ubuntu 19.10 was used due to its recent software components being at similar versions to Intel’s rolling-release distribution.
Besides Ubuntu 19.10 out-of-the-box, the video encoder tests were done on the i9-9900K as well when setting to the P-State performance governor (the default of Clear Linux) and also when using the same CFLAGS/CXXFLAGS as Clear Linux that it applies by default.
With SVT-AV1, Ubuntu was just about a frame behind Clear Linux though I’ll also repeat these tests on some Xeon servers where in the past at least the difference is more meaningful.
The speedier SVT-HEVC encoder though does illustrate a 5% advantage to Clear Linux. But even when changing to the performance governor or matching the compiler flags, Clear was still carrying a 5% lead.
SVT-VP9 was also maintaining a slight lead on Clear Linux over Ubuntu 19.10.
Long story short, simply copying the compiler flags and using the same performance governor of Clear Linux does not account for the video encode advantages to the Intel open-source distribution, contrary to the belief of some within our forums and elsewhere. Some additional tests still ongoing for an interesting follow-up shortly.
>> Source Link